Thus far, the capabilities of the top AI models were hidden away in arcane benchmarks which didn’t interest anyone outside the field, but there might now be ways for people outside of tech of see these models in action.
Data Science competition platform Kaggle has launched Game Arena, which is an AI chess exhibition tournament. The tournament will feature models including DeepSeek R1, OpenAI’s o3 and o4-mini, Google’s Gemini 2.5 Pro and Gemini 2.5 Flash, Claude’s Opus 4, xAI’s Grok 4 and Moonshot AI’s Kimi-2. These AI models will play chess with each other, and the matches will be livestreamed using visualizers so that people can follow the action. Google DeepMind is partnering with Kaggle for the event.
“Games are an excellent foundation for robust AI evaluation that helps us understand what really works (and what doesn’t) against complex reasoning tasks,” Kaggle said. “Their structured nature and measurable outcomes make them the perfect testbed for evaluating models,” it added.
And Kaggle has managed to rope in some chess celebrities for the tournament. The most popular chess streamer on YouTube and current world number 2 Hikaru Nakamura will be doing commentary for the event. Levy Rozman, another popular chess content creator will be doing daily recaps for the games. The games will also be streamed on Take Take Take, the chess app with which World number 1 Magnus Carlsen is involved.
A chess tournament with top AI models could be quite interesting to watch. Thus far, models only competed against each other on benchmarks, and those scores were the only way to gauge their abilities. But seeing these models compete with one another in a game that requires high degrees of thinking and intelligence could not only be interesting to watch, but also give people some insights into their relative abilities. Granted, chess skills are only a small subset of what an AI model is capable of, but they could be a fairly robust reflection of the general thinking and reasoning abilities of the model, and this Kaggle tournament would be an opportunity for the model providers to make a splash with their creations among the general public.